Seed Dispersal by Animals: Exact Identification of Source Trees with Endocarp DNA

Seed dispersal by animals: exact identification of source trees with endocarp DNA microsatellites

JOSÉ A. GODOY* and PEDRO JORDANO † *Laboratorio de Ecología Molecular, Estación Biológica de Doñana, CSIC, Apdo. 1056, E-41080 Sevilla, Spain, †Departamento de Biología Evolutiva, Estación Biológica de Doñana, CSIC, Apdo. 1056, E-41080 Sevilla, Spain

Abstract A long-standing challenge in studies of seed dispersal by animal frugivores has been the characterization of the spatial relationships between dispersed seeds and the maternal plants, i.e. the seed shadow. The difficulties to track unambiguously the origin of frugivore- dispersed seeds in natural communities has been considered an unavoidable limitation of the research field and precluded a robust analysis of the direct consequences of zoochory. Here we report that the multilocus genotype at simple sequence repeat (SSR; microsatellite) loci of the woody endocarp, a tissue of maternal origin, provides an unequivocal genetic fingerprint of the source tree. By comparing the endocarp genotype against the complete set of genotypes of reproductive trees in the population, we could unambiguously identify the source tree for 82.1% of the seeds collected in seed traps and hypothesize that the remaining 17.9% of sampled seeds come from other populations. Identification of the source tree for Prunus mahaleb seeds dispersed by frugivores revealed a marked heterogeneity in the genetic composition of the seed rain in different microhabitats, with a range of 1– 5 distinct maternal trees contributing seeds to a particular landscape patch. Within- population dispersal distances ranged between 0 and 316 m, with up to 62% of the seeds delivered within 15 m of the source trees. Long distance dispersal events, detected by the exclusion of all reproductive trees in the population, accounted for up to 17.9% of the seeds sampled. Our results indicate strong distance limitation of seed delivery combined with infrequent long-distance dispersal events, extreme heterogeneity in the landscape pattern of genetic makeup, and a marked mosaic of multiple parentage for the seeds delivered to a particular patch. Keywords: avian frugivores, microsatellites, seed dispersal, seed endocarp, seed shadow, spatial variation

The difficulties in understanding, measuring and Introduction analysing dispersal of seeds in natural communities has For animal-dispersed, endozoochorous, plant species a been considered as an unavoidable limitation of the seed shadow is the population recruitment surface ( research field (Wheelwright & Orians 1982), and the Janzen methodological shortcomings of tracking the origin of 1970; Connell 1971), the spatial pattern of seed frugivore-dispersed seeds has precluded a deep and distribution over the landscape relative to parent trees robust analysis of the direct consequences of zoochory and other conspecifics that results from the process of (Levey & Sargent 2000). seed dispersal. Frugivores determine the landscape Seed and pollen are the two vectors of gene flow in pattern of seed distribution and density over available plants. Levels of gene flow have traditionally been microhabitats and thus can have a dramatic effect on both assessed through indirect methods that infer average the demography and genetic make-up of animal- historical values from the extent of subpopulation dispersed plant species. differentiation under a particular model (Neigel 1997; Sork et al. 1999). Within population gene movement has Correspondence: Pedro Jordano. Fax: +34 954621125; E-mail: been addressed by analysing the genetic structuring at [email protected] a microgeographic scale 228 J. A. GODO Y and P. J ORD AN O

(Alvarez-Buylla et al. 1996; Epperson & Alvarez-Buylla southeastern Spain) in a Prunus mahaleb population located 1997; Aldrich et al. 1998; see also Ueno et al. 2000). A in Nava de las Correhuelas (1615 m elevation) (Valle et al. prevailing pattern in these studies is strong clumping of 1989). Frugivorous birds and mammals visiting P. mahaleb dispersed seeds, significant autocorrelation values over trees in Spanish populations usually behave as legitimate short distances, and extreme context-sensitivity of the seed dispersers swallowing pattern of genetic structuring depending on the particular landscape setting. More recently, the availability of hypervariable genetic markers (Tautz 1989) has allowed the direct assessment of contemporary gene flow in a landscape context, based on parentage exclusion analyses (Luikart & England 1999). Direct analyses in plants have concentrated on paternity assessment for progeny arrays of known maternity with two main objectives: analysis of male fertility (Meagher 1986, 1991), and estimation of pollen-mediated gene flow (see e.g. Devlin et al. 1988; Devlin & Ellstrand 1990). How- ever, very few papers so far have applied a direct approach to the study of seed dispersal, the second component of gene flow. This is in part due to the limited resolution of parental exclusion methods, especially when the two parents are unknown, and the impossibility of distinguishing the father and mother in cosexual species (such as hermaphro- dites or monoecious; Dow & Ashley 1996; Schnabel et al. 1998a). Cytoplasmic markers could also be used to exclude maternity but they usually lack enough variation for individual maternity assignments (Ouborg et al. 1999). Here we report a new approach for the identification of the source tree of individual Prunus mahaleb seeds recovered from animal frugivores (either defecated or regurgitated) based on analysis of simple sequence repeats (SSRs or microsatellites) (Tautz 1989) from the seed endocarp DNA. The multilocus genotype of the seed endocarp unambiguously identifies the maternal tree of the seed due to the fact that the endocarp tissue in Prunus species, as in most drupaceous species, is of maternal origin, derived from the carpelar wall (Roth 1977). We determined the minimum number of female trees contributing progeny to specific locations in the landscape and the fraction of own progeny dispersed beneath a given source tree relative to progeny from other neighbouring trees. In addition, we analysed the spatial relationships between individual dispersed seeds and their source tree by determining seed dispersal distances directly.

Materials and methods

Study site and field methods This study was conducted in the Reserva de Navahondona- Guadahornillos (Parque Natural de las Sierras de Cazorla, Segura y las Villas, Jaén province, TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 229 the fruits whole and defecating and/or regurgitating the centrifuged for 15 min at 16 000 g. The pellet was washed seeds, usually after leaving the tree ( Jordano 1994). Most with 70% ethanol, dried and resuspended in 100 μL TE seed rain of P. mahaleb in the study areas is contributed by buffer. frugivorous birds. Seed rain and the resulting recruitment pattern of seedlings and saplings are highly patchy, and largely restricted to microhabitats beneath woody cover in the vicinity of fruiting trees ( Jordano & Schupp 2000).

Sampling strategy Leaf tissue was sampled for total of 180 adult trees, representing 100% of the reproductive population, during the years 1996 – 2000 ( Jordano & Godoy 2000). To compare the maternal genotype obtained from leaf tissue with the genotype of the seed endocarp, we sampled 1– 3 seeds from the branches of four trees. In addition, we obtained the endocarp and embryo tissues from 11 seeds among the progeny (five families) obtained in diallel crosses from a concurrent study (P. Jordano, personal observation); thus, for these seeds both the sire and dam trees were known. Naturally dispersed seeds were sampled in the dispersal season of 1996 with 20 replicate sets of two seed traps located in four different microhabitat types in the forest: beneath five focal P. mahaleb trees, beneath 10 mid-height shrubs (e.g. Crataegus mongyna, Lonicera arborea), beneath three pine trees, and beneath two pine trees with juniper understory. These replicate sampling points were a stratified random sample of a total of 481 sampling points used in a concurrent study of seed rain patterns. The types of microhabitats were a subset of those defined in Jordano & Schupp (2000). The adult reproductive trees in the population were mapped and their locations recorded in a GIS database, which also included the locations of the sampling sites with seed traps. A total of 95 seeds were randomly subsampled from those recovered from the traps, either in defecations or regurgitations of bird and mammal frugivores.

DNA extraction and microsatellite genotyping DNA was extracted from 100 to 200 mg of fresh leaf tissue, sampled in 1998 – 99, using the rapid miniprep method of Cheung et al. (1993). Briefly, tissue was homogenized in 320 μL of extraction buffer (200 mm Tris-HCl pH 8.0, 70 mm EDTA, 2 m NaCl, 20 mm sodium bisulfite) with an electric drill (560 W; full speed) with attached plastic disposable pestles. After homogenization 80 μL of 5% sarcosyl was added and the sample was incubated at 65 °C for 30 min and centrifuged at 16 000 g for 15 min to remove insoluble material. DNA was precipitated by the addition of 90 μL of 10 m ammonium acetate and 200 μL of isopropanol. The mixture was incubated at room temperature for 5 min and 230 J. A. GODO Y and P. J ORD AN O

The seed endocarp was split open and separated by Source tree identification hand from the seed content (embryo plus vestigial For the study of seed dispersal, a total of 182 adult trees endosperm), and collected for later grinding. We do not and expect contamination of the endocarp tissue with remains 95 dispersed seed endocarps were genotyped. The source of the seed content, as the whole embryo is separated tree for individual dispersed seeds was identified by from the endocarp by a thin seed coat that is cleanly comparing the endocarp multilocus genotype with the extracted when the endocarp is split open. For the DNA extraction from seed endocarps we used the following modifications of the protocol for leaves: tissue was previously grinded in a mixer-mill and homogenized in 320 μL of extraction buffer, and the DNA finally resuspended in 50 μL TLE (200 mm Tris-HCl pH 8.0, 70 mm EDTA). DNA yields were estimated fluorometrically for leaf extracts, but they were too low (range 1– 5 ng/μL) to be accurately estimated for endocarp extracts. For the endocarps we empirically tested a series of extract volumes in amplification reactions of microsatellite marker and found that 5 μL (1/10) of endocarp extract yielded consistent amplifications. We tested a total of 43 primers pairs designed for cultiv- ated Prunus and Malus species (A. Abbott 1998, personal communication; G. King 1998, personal communication; Cipriani et al. 1999; Downey & Iezzoni 2000; Sosinski et al. 2000) and selected a subset of nine markers that showed polymorphism in P. mahaleb for use in this study (Table 1). As a template for the polymerase chain reaction (PCR) we used 30 ng of leaf DNA or 5 μL (1/10) of endocarp extract. PCR was performed in a final volume of 20 μL

con- taining 67 mm Tris-HCl pH 8.8, 16 mm (NH4)2SO4, 2 mm MgCl2, 0.01% Tween-20, 0.01% BSA, 0.25 mm of each dNTP, 0.25 μm of each primer, and 0.5 U of Taq DNA polymerase. Reactions were incubated in a MJ Research PTC-100 thermocycler programmed for a ‘touchdown’ PCR as follows: an initial denaturation step at 94 °C for 2 min; 17 cycles of 92 °C for 30 s, annealing at 66 – 50 °C for 30 s (1 °C decrease in each cycle), and extension at 72 °C for 30 s; 19 cycles of 92 °C for 30 s, 50 °C for 30 s, and 72 °C for 30 s. A final extension was programmed at 72 °C for 5 min. Amplified fragments were analysed using an ABI 377 Genetic Analyser (Applied Biosystems). To assess the genotypic relationships among the leaf tissue, the endocarp of the seed, and the embryo, progeny from a diallel cross of known sires and dams were obtained in 1992 and 1993 and independent DNA extractions and genotypes were obtained from each tissue. When comparing embryo to parental genotypes, a strict Mendelian inheritance was observed with no evidence for the occurrence of null alleles. TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 231 complete set of genotypes of reproductive trees in the population. For each sampled seed, the adult individual having a genotype matching the seed endocarp genotype was assigned as the mother tree. Matches between endocarp and adult genotypes were found by examining all possible pairwise comparisons between endocarps and adult trees. A hypothesis of identity (rp = 1, rm = 1) was tested with program kinship (version 5.0.5) and the significance estimated by a jackknife resampling method (Queller & Goodnight 1989). Briefly, a hypothesis about pedigree identity between an endocarp and a tree (leaf tissue) is defined in terms of rp and rm , the probabilities that individuals in the pair share an allele by direct descent from their father or mother, respectively. For the hypothesized relationship of full genotypic matching, both r-values would be 1.0 (K.F. Goodnight, personal communication). Given the hypothesis, the r-values, the population allele frequencies, and the two genotypes under consideration are used to estimate the likelihood that this genotype combination could have been produced by the relationship as specified. The method calculates a likelihood for two such hypotheses, the primary hypothesis (identity) and a null hypothesis (no identity), and the ratio between them (primary/null). A high value of the ratio favours the primary hypothesis and a low value rejects it in favour of the null hypothesis. The significance level of a given ratio is estimated empirically, by simulation, generating pairs of individuals using the hypothesis settings and the allele frequencies and determines the ratio needed to reject the null hypothesis with a given P < 0.05 (Queller & Goodnight 1989).

Results

Genotyping of seed endocarp tissue for the identification of the source (maternal) tree We successfully extracted DNA from the woody endocarp tissue of individual Prunus seeds and complete microsatellite genotypes were thus obtained. Reliable genotypes were obtained from endocarp tissue, both from freshly collected seeds and from seeds collected in seed traps. Most seeds sampled in seed traps were dispersed by birds, small to medium-sized passerines, with short gut passage times that either defecate or regurgitate them ( Jordano & Schupp 2000). In addition, we have also amplified successfully extracted DNA from endocarps of seeds dispersed by red foxes (Vulpes vulpes), stone martens (Martes foina) and large birds such as wood pigeon (Columba palumbus) and carrion crow (Corvus corone). Thus, the miniprep DNA extraction protocol can be used to obtain adequate amounts of DNA even from seeds which have passed the digestive tract of frugivores. Because endocarp tissue is of maternal origin its multilocus genotype should be identical to that of the mother 232 J. A. GODO Y and P. J ORD AN O

Table 1 Multilocus genotypes for leaf, endocarp, and embryos of Prunus mahaleb determined from scoring of simple sequence repeat (SSR) loci. For each maternal tree, the presence of alleles for each of nine SSR loci is given for the leaf tissue and for seeds of its progeny. Two or three seeds were selected from each tree and the allelic profile for the SSR loci is given for both the endocarp tissue (labelled with T) and the embryo (labelled E). Empty boxes indicate those instances (all correspond to embryos) where a particular allele is not present in the mother tree. Endocarps have in all instances identical multilocus genotypes to the leaf tissue. For example, embryo 1926 5E is homozygous for locus pchgms3 (191/191), while both the leaf tissue (1926) and the endocarp tissue (1926 5T) are heterozygous for this locus (179/191). Alleles are designed by the size (bp) of their products

TREE 1926 TREE 1927

Endocarps Embryos Endocarps Embryos Leaves Leaves

Locus Allele 1926 1926 5T 1926 6T 1926 5E 1926 6E 1927 1927 1T 1927 4T 1927 1E 1927 4E

UDP96-001* 114 122 124 pchgms3† 179 191 UDP96-018* 246 248 UDP97-403* 105 107 109 PS12A02† 171 175 185 pchcms5† 235 239 UDP98-406* 98 102 UDP97-402* 140 148 152 MS01A05‡ 196 200

TREE 1928 TREE 1847

Endocarps Embryos Endocarps Embryos Leaves Leaves

Locus Allele 1928 1928 1T 1928 3T 1928 4T 1928 1E 1928 3E 1928 4E 1847 1847 2T 1847 3T 1847 2E 1847 3E

UDP96-001* 122 124 126 pchgms3† 179 187 195 UDP96-018* 248 UDP97-403* 107 109 125 145 PS12A02† 185 189 pchcms5† 235 239 UDP98-406* 98 100 UDP97-402* 142 148 MS01A05‡ 200 202

TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 233 *Cipriani et al. (1999); †Sosinski et al. (2000); ‡G. King, personal communication. 234 J. A. GODO Y and P. J ORD AN O tree. We confirmed this for individual mother trees by contributing progeny to comparing the genotypes of leaf and tissue from their seed progeny (both endocarp and embryo) (Table 1). We further tested this hypothesis by analysing 1– 5 seeds per family obtained in a series of diallel crosses; for each of these seeds both the embryo and the endocarp were genotyped and compared to those of the known dam and sire trees (data not shown). As expected, the embryo genotypes were vari- able and compatible with the genotypes of the known sire and dam trees. On the other hand, the endocarp genotypes of seeds collected from the same tree were identical to each other and to the genotypes obtained from the leaf tissue of their corresponding maternal trees (Table 1). The genotyping of seed endocarp thus provides a convenient tool for the unambiguous assignment of maternity for seeds dispersed by frugivorous animals.

Identification of the maternal source trees for dispersed seeds We genotyped a total of 182 reproductive trees, ~100% of the potential maternal trees for the 1996 seed crop, and 95 dispersed seeds collected in seed traps in different microhabitats. In this population we determined a total of 182 distinct multilocus genotypes, each adult tree in the population showing a unique multilocus genotype. Gene diversity was 0.996 ± 0.001 (Nei 1987), with mean expected heterozygosity of 0.586 and an average of 6.7 alleles per locus (estimated with the arlequin package, version 2.0; Scheiner et al. 2000). For 78 seeds in the seed traps (82.1%) a significant genotype match was found with a single adult tree in the population. In these cases, the matching tree was identified as the maternal source tree of the seed. No matching genotypes were found among adult trees in the population for the remaining 17 seeds, indicating that they come from maternal trees located in other populations. This result reveals the occurrence of long-distance interpopulation seed dispersal in this system.

Analysis of the seed shadow A typical seed shadow of Prunus mahaleb in our study area consists of most seeds dispersed beneath the canopy of the maternal trees or beneath cover of mid-height shrubs (Crataegus monogyna, Lonicera arborea, Rosa spp.), with much lower seed fall figures beneath low shrubs ( Juniperus spp., Berberis vulgaris) or pine trees. The identification of the source tree for individual dispersed seeds allowed us to characterize the composition and structure of the seed shadow in the population. Differ- ent microhabitats within the landscape were sampled for dispersed seeds and their tree of origin was identified. The mean number of maternal trees TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 235 any seed sampling location ranged between 1 and 5 a (Table 2), and this number differed significantly among microhabitat types (F = 4.78, P = 0.007, d.f. = 3, 34). Sampling sites beneath Prunus and mid-height shrubs had a greater number of maternal trees contributing seed than sites beneath pine trees (Scheffé a posteriori test, P < 0.04 for both comparisons). Most of the seeds dispersed beneath a P. mahaleb tree were from its own progeny (Table 2, Fig. 1), although trees differed in the proportion of their own progeny repres- ented in the seed rain beneath them. Thus, individual seeds dispersed beneath fruiting trees originated from a varied number of conspecifics. The sampling points located beneath mid-height shrubs close to a P. mahaleb tree typically showed a large proportion of the seed shadow composed by progeny from the nearest tree (the ‘focal’ tree; Fig. 1). The relative representation of progeny from the ‘ focal’ tree dramatically decreased with increasing distance, and was very low beneath pine trees growing at the forest edge, away from the shrub patches where most P. mahaleb source trees are located. This indicates not only a potentially strong distance limitation, but also extreme clumping or aggregation of different seed genotypes.

Direct estimation of dispersal distances The fact that both the maternal trees (seed sources) and the sampling points with seed traps were georeferenced allowed a direct estimation of seed dispersal distances, even with this limited sample of dispersed seed (Fig. 2). As expected from observational study of flight patterns and frugivore foraging (Jordano & Schupp 2000), most dispersal distances were short, within 10 m of the maternal source tree. The median dispersal distance for this seed sample was 6.1 m (0.0 – 46.1 m) (median and 25 – 75% per- centiles), with an extremely skewed frequency distribution (Fig. 2), and approximately 65% of the dispersal distances being < 20 m. These empirically derived estimates correlate well with our previous characterization of the seed shadow based on direct watches of bird foraging (Jordano & Schupp 2000). Interestingly, the dispersal distances estimated for seeds trapped in different microhabitats differed significantly (H = 32.81, P < 0.0001; Kruskall– Wallis test). Seeds dispersed beneath pine trees had longer dispersal distances than those sampled beneath P. mahaleb trees, mid-height shrubs, or pine trees with low shrub understory.

Discussion The endocarp tissue from Prunus mahaleb seeds dispersed by frugivores, that either regurgitate or defecate them, can be used to identify the maternal source tree when their genotypes are compared with genotypes of adult trees in 236 J. A. GODO Y and P. J ORD AN O

Table 2 Assignment of maternal trees for animal-dispersed, Prunus mahaleb seeds sampled in seed traps

Seed Other trees Number of Number of seeds Estimated number sampling Focal contributing seeds identified trees with maternity of trees contributing point* tree* to sampling point* contributing seeds unassigned seeds†

Beneath Prunus mahaleb 795 3725 19281 2 1 3 793 19273 18462 18371, 19211 4 0 4 799 3996 — 1 2 3 792 18436 — 1 0 1 791 19216 18231, 18341 3 1 4 Beneath mid-height shrubs 830 19270 3721, 3641, 3701 3 2 5 831 19271 3963, 19251, 19241, 3971 5 0 5 832 19270 18491 1 4 4 833 19270 3791, 3831 2 0 2 826a 19214 19251 2 1 3 826b 18431 — 1 0 1 827 — 19232, 11641, 18311 3 3 5 828 — 3611, 3621, 19271, 3961, 18451 5 0 5 837 3993 19272 2 1 3 838 3993 19271 2 1 3 Beneath pine trees 866 3990 19391, 3641 2 0 2 867 3990 18481, 11881 2 0 2 846 3990 19321 1 0 1 847 3990 11901 1 1 2 849 3990 11791 1 0 1

*Numbers refer to the codes of the sampling points or the tree codes. Superscripts indicate the number of seeds that were assigned to each maternal tree. Each of the ‘focal’ trees had a set of seed traps associated. †Figures include the number of identified maternal trees plus the estimated number of trees contributing the unassigned seeds. Thus, whenever two unassigned seeds had mismatching genotypes we computed two distinct (unassigned) maternal trees. population. This is expected on the basis of the anatomical origin of Prunus drupes, with a fleshy mesocarp and a woody endocarp derived from the diploid carpelar tissues of the mother flower (Roth 1977). For dispersed seeds that still retain pieces of pulp, the maternal source tree could also be identified by genotyping pulp remains. However, using seeds is preferable as the pulp rapidly molds and disappears from fruit remains sampled with seed traps. Thus, this method of direct genotyping can be easily combined with regular sampling schemes of seed rain using seed traps (Kollmann & Goetze 1997; Harms et al. 2000) to assess patterns of seed dispersal at the landscape level. Estimating seed dispersal distances based on seed traps can be effective depending on the robustness of the seed trap sampling scheme. Ideally, seed traps should be arranged randomly throughout the habitat, in a stratified scheme according to major landscape units, and then a sufficient number of the seeds sampled in the traps should be selected for genotyping. We are working with such an intensive sampling scheme, with a network of 1400 seed traps in two populations, to assess the robustness and potential biases of direct estimates of dispersal TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 237 distances ( J.L. García-Castaño, J.A. Godoy & P. Jordano, in preparation). The approach we describe here can be applied to a variety of endozoochorous seed species that typically show a thick endocarp, although care should be taken in determining the anatomical origin of the tissue analysed. When assessing species with complex fruit structures, such as arillate seeds, etc., a preliminary comparison with other maternal tissues can be undertaken to assure the reliability of using a particular tissue to compare with the maternal genotype. The method can also be used with wind- dispersed species that typically show ancillary structures such as wings, pappus, etc., presumably of maternal origin. Our approach can be used to estimate relative female fertilities, the diversity of trees contributing seeds to particular landscape patches and thus, the heterogeneity of the seed rain over the landscape, especially in relatively small populations. Some of these parameters can be estimated even with incomplete genotyping of the adults in larger populations (Slate et al. 2000). For instance, an exclusion approach requiring limited genotyping effort can be used to test whether a subset of candidate trees in TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 233

6 conspecifics. A heterogeneous landscape, with a marked

Other 5 Own s d

e 4 e s

3 r e b

m 2 u N

0 0 <10 >10 Distance to nearest Prunus tree (m)

Fig. 1 Type of progeny of animal-dispersed Prunus mahaleb seeds sampled in seed traps as a function of distance to the nearest P. mahaleb tree. ‘Own’ refers to seeds from the nearest tree, i.e. closest to the seed trap locations in the forest. ‘Other ’ refers to those seeds whose maternal tree was not the nearest tree.

40 t n e c

30 r e P 20

0 0 50 100 150 200 250 300 350 400 Distance from maternal tree (m) Fig. 2 Frequency distribution of seed dispersal distances estimated by direct genotyping of seed endocarp of animal- dispersed Prunus mahaleb seeds sampled in seed traps. the neighbourhood of a given seed sampling point are the source for the sampled seeds. Sampling points beneath P. mahaleb, mid-height shrubs, or pines with low shrub understory recruited seed rain from a larger number of parents than points beneath pines. Thus, those patches in the habitat with dense vegetation cover not only recruited larger numbers of seeds ( Jordano & Schupp 2000) but also showed greater small-scale genetic diversity (estimated from the number of distinct maternal trees contributing progeny). The fact that P. mahaleb trees typically grow associated to patches with high shrub cover results in each maternal tree acting as a dispersal site, or ‘sink’, for seeds dispersed from other 234 J. A. GODO Y and P. J ORD AN O mosaic of covered and open microhabitats, and the nonrandom movements of frugivores strongly selecting covered microhabitats (Kollmann & Schneider 1996; Jordano & Schupp 2000) result in a dramatic heterogeneity in the provenance of the seeds and a marked clumping of the distinct seed genotypes at different sites over the landscape. This has important implications in conservation and management of forest species: simply adding a low shrub cover beneath pine trees would increase significantly both the density of dispersed seeds and its genetic heterogeneity. Whether or not this translates into actual recruitment of seedlings and saplings would depend on factors acting during the postdispersal stages of regeneration. Our empirically derived estimates of dispersal distances reveal an extremely skewed frequency distribution, with most seeds dispersed in the neighbourhood of the maternal tree. These estimates, relying on seed trap sampling, do not account for the fact that secondary dispersal after delivery by frugivores might occur. This is not a major influence in our system, as secondary dispersal by ants is infrequent (P. Jordano, personal observation), although seed caching by Apodemus sylvaticus has been reported (Hulme 1997; E.W. Schupp and P. Jordano, personal observation). Despite the fact that no less than 60% of seed dispersal events were within 20 m of the maternal source tree, we recorded ~10% of dispersal distances > 100 m away. These were consistently seeds dispersed beneath pine trees, generally located along the forest edge surrounding the deep soil ‘poldje’ patches and lower rocky slopes where P. mahaleb trees grow. Thus, any P. mahaleb tree has a relatively long distance to pine trees in this population. We have previously reported ( Jordano & Schupp 2000) that the seed rain to pine trees is largely contributed in our study area by Turdus viscivorus, a thrush species that typically flies to perch on pine trees after feeding on P. mahaleb. In contrast, the seed rain reaching covered microhabitats (beneath P. mahaleb, mid- height shrubs and patches with low shrubs) is contributed by an assortment of up to six small-bodied frugivore species (warblers, robin, and redstarts) with short flight distances. Therefore our results, in combination with previous field observations, indicate that most long-distance dispersal in this population is directed to a particular landscape patch despite a relatively diverse frugivore assemblage; and those long- distance dispersal events are contributed by a single disperser species. All our estimates of dispersal distances do not include the very long dispersal distances from other populations (immigrant seeds), so they should be taken as a within- population dispersal estimate. At least 17 sampled seeds (17.9%) were not assignable to any adult maternal tree in TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 235 the populatio6 n and this can be taken as an upper estimate conspecifics. A heterogeneous landscape, with a marked of the fraction of immigrant seeds in the seed rain. This result reveals the occurrence of long-distance interpopulation seed dispersal in P. mahaleb, presumably from surrounding populations which are 3 – 5 km away. Evidence for extensive gene flow in the fragmented P. mahaleb populations has been reported, with high levels of within- population genetic diversity estimated from random amplified polymorphic DNA (RAPD) markers ( Jordano & Godoy 2000). We are currently working with material from the two nearest populations to be able to identify the source population for unassigned seeds. Dispersal distances estimations add evidence to our previous finding of a strong spatial autocorrelation in the 0 – 20 m distance for the coancestry estimates among adult trees in this population based on SSR and RAPD markers ( Jordano & Godoy 2001). Despite the fact that there are a number of multiple postdispersal influences determining survival of seeds and seedlings to adulthood that can erase the initial spatial pattern determined by frugivore activity ( Jordano & Herrera 1995; Rey & Alcántara 2000), it seems that patterns of genetic structuring in this population track the influences of seed delivery by frugivores. This finding also reveals that fragmentation of populations of animal- dispersed woody species can result in severe genetic bottlenecks by isolating patches of genetically similar adult trees (Loiselle et al. 1995; Nason & Hamrick 1997; Schnabel et al. 1998b). Also, the extinction or marked seasonal or annual changes in certain frugivore species can severely alter the spatial patterns of genetic make-up in the seed shadow by truncating and/or redirecting gene flow via seeds. We suggest that a distance limited, spatially aggregated pattern in the genetic make up of the seed shadow, as reported here for P. mahaleb, is a typical situation of many animal-dispersed, fleshy-fruited plant species that deserves further study. Our analysis also indicates that relatively long dispersal distances are probably not infrequent, but represent an extremely small fraction of the dispersal events in a given reproductive episode, and are extremely directed in the heterogeneous landscape, as the result of nonrandom foraging by a limited set of species in the disperser assemblage. The combined application of molecular tools based on hypervariable SSR markers and detailed observational and experimental ecological data has allowed a robust characterization of the influence of frugivore foraging on P. mahaleb population dynamics. This pervades the demographic effects — frequently the main and unique subject of ecological studies — to show up as a multiple, combined, demographic and genetic influence. To our knowledge this is the first study to combine detailed ecological and molecular approaches to reveal the origin and spatial patterns of the genotypes of animal-dispersed 236 J. A. GODO Y and P. J ORD AN O seeds. Here we have reported the results of a limited of sampling design to reveal the potentials and weaknesses a refined method for estimating gene flow from angiosperm paternity analysis. Evolution, 44, 248 – 259. of our approach, and more in depth study is under way. The combination of carefully designed ecological experiments and sound laboratory analyses with molecular markers pave a promising avenue for future studies of plant– frugivore mutualisms.

Acknowledgements

We are indebted to Alicia Prieto, Rocío Requerey, and especially Manolo Carrión, Juan Luis García-Castaño, Juan Miguel Arroyo, Myriam Márquez, Nieves Capote, Jesús Rodríguez, and Joaquín Muñoz for generous help with field and laboratory work and making this study possible. The help, advice, and discussions with Gene W. Schupp and Janis Boettinger were fundamental during many stages of this project. Birgit Ziegenhagen, Pere Arús, and Rémy Petit provided advice and useful discussion on DNA extraction protocols and related issues. We feel especially indebted to Albert Abbott, Graham King and Amy Iezzoni, who generously provided Prunus microsatellites before publication. The Agencia de Medio Ambiente, Junta de Andalucía, provided generous facilities that made this study possible in the Sierra de Cazorla and authorized our work there. The early stages of this research were impelled by the generous advice of Juan Jordano, Concha Almoguera and Pere Arús; we appreciate their support and technical advice that made this project possible. We thank Jordi Bascompte for useful discussion and comments on an early draft of the manuscript. This work was supported by grants 1FD97- 0743-C03-01 and BOS 2000-1366-C02-01 from the Comisión Interministerial de Ciencia y Tecnología, Ministerio de Educación y Ciencia and the European Commission, and also by funds from the Consejería de Educación y Ciencia, Junta de Andalucía.

References

Aldrich PR, Hamrick JL, Chavarriaga P, Kochert G (1998) Micro- satellite analysis of demographic genetic structure in fragmented populations of the tropical tree Symphonia globulifera. Molecular Ecology, 7, 933 – 944. Alvarez-Buylla ER, Chaos A, Piñero D, Garay AA (1996) Demo- graphic genetics of a pioneer tropical tree species: patch dynamics, seed dispersal, and seed banks. Evolution, 50, 1155 – 1166. Cheung WY, Hubert N, Landry BS (1993) A simple and rapid DNA microextraction method for plant, animal, and insect suit- able for RAPD and PCR analyses. PCR Methods and Applications, 3, 69 – 70. Cipriani G, Lot G, Huang WG, Marrazzo MT, Peterlunger E, Testolin R (1999) AC/GT and AG/CT microsatellite repeats in peach [Prunus persica (L. ) Batsch]: isolation, characterisation and cross-species amplification in Prunus. Theoretical and Applied Genetics, 99, 65 – 72. Connell JH (1971) On the role of natural enemies in preventing competitive exclusion in some marine animals and in rain forest trees. In: Dynamics of Populations (eds Den Boer PJ, Gradwell G), pp. 298 – 312. Centre for Agricultural publishing and Docu- mentation (PUDOC), Wageningen, The Netherlands. Devlin B, Ellstrand NC (1990) The development and application TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 235

Devlin B, Roeder K, Ellstrand NC (1988) Fractional paternity Neigel JE (1997) A comparison of alternative strategies for assignement: theoretical development and comparison to other estimat-tion of Chamaelirium luteum. II. Patterns of male reproductive methods. Theoretical and Applied Genetics, 76, 369 – 380. success. American Naturalist, 137, 738 – 752. Dow BD, Ashley MV (1996) Microsatellite analysis of seed Nason JD, Hamrick JL (1997) Reproductive and genetic con- dispersal sequences of forest fragmentation: two case studies of and parentage of saplings in bur oak, Quercus macrocarpa. neotropical Molecular Ecology, 5, 615 – 627. canopy species. Journal of Heredity, 88, 264 – 276. Downey SL, Iezzoni AF (2000) Polymorphic DNA markers in Nei M (1987) Molecular Evolutionary Genetics. Columbia University black cherry (Prunus serotina) are identified using sequences Press, New York, US. from sweet cherry, peach, and sour cherry. Journal of the American Society of Horticultural Science, 125, 76 – 80. Epperson BK, Alvarez-Buylla ER (1997) Limited seed dispersal and genetic structure in life stages of Cecropia obtusifolia. Evolu- tion, 51, 275 – 282. Harms KE, Wright SJ, Calderón O, Hernández A, Herre EA (2000) Pervasive density-dependent recruitment enhances seedling diversity in a tropical forest. Nature, 404, 493 – 495. Hulme PE (1997) Post-dispersal seed predation and the establishment of vertebrate dispersed plants in Mediterranean scrublands. Oecologia, 111, 91– 98. Janzen DH (1970) Herbivores and the number of tree species in tropical forests. American Naturalist, 104, 501– 528. Jordano P (1994) Spatial and temporal variation in the avian- frugivore assemblage of Prunus mahaleb: patterns and consequences. Oikos, 71, 479 – 491. Jordano P, Godoy JA (2000) RAPD variation and population genetic structure in Prunus mahaleb (Rosaceae), an animal-dispersed tree. Molecular Ecology, 9, 1293 –1305. Jordano P, Godoy JA (2001) The dynamics of frugivore-generated seed shadows: demographic and genetic effects. In: Frugivores and Seed Dispersal: Ecological, Evolutionary, and Conservation Issues (eds Levey DJ, Silva W, Galetti M), in press. Commonwealth Agricultural Bureau, Wallingford, England. Jordano P, Herrera CM (1995) Shuffling the offspring: uncoupling and spatial discordance of multiple stages in vertebrate seed dispersal. Ecoscience, 2, 230 – 237. Jordano P, Schupp EW (2000) Determinants of seed dispersal effectiveness: the quantity component in the Prunus mahaleb– frugivorous bird interaction. Ecological Monographs, 70, 591– 615. Kollmann J, Goetze D (1997) Notes on seed traps in terrestrial plant communities. Flora, 192, 1–10. Kollmann J, Schneider B (1996) Effects of landscape structure on seed dispersal of fleshy-fruited species along forest edges. Bulletin of the Geobotanical Institute, 63, 77 – 86. Levey DJ, Sargent S (2000) A simple method for tracking vertebrate- dispersed seeds. Ecology, 81, 267 – 274. Loiselle B, Sork VL, Nason J, Graham C (1995) Spatial genetic structure of a tropical understory shrub, Psychotria officinalis (Rubiaceae). American Journal of Botany, 82, 14209 –11425. Luikart G, England PR (1999) Statistical analysis of microsatellite DNA data. Trends in Ecology and Evolution, 14, 253 – 256. Meagher TR (1986) Analysis of paternity within a natural population of Chamaelirium luteum. I. Identification of the most-likely male parents. American Naturalist, 128, 199 – 215. Meagher TR (1991) Analysis of paternity within a natural popula- TR AC KI N G TH E S OU RC E O F AN IM AL - DI SPER SE D S EE D S 236

Devlin B, Roeder K, Ellstrand NC (1988) Fractional paternity Neigel JE (1997) A comparison of alternative strategies for ing gene flow from genetic markers. Annual Review of Ecology Doñana.estimat- and Systematics, 28, 105 –128. Ouborg NJ, Piquot Y, van Groenendael JM (1999) Population genetics, molecular markers and the study of dispersal in plants. Journal of Ecology, 87, 551– 568. Queller DC, Goodnight KF (1989) Estimating relatedness using genetic markers. Evolution, 43, 258 – 275. Rey PJ, Alcántara JM (2000) Recruitment dynamics of a fleshy- fruited plant (Olea europaea): connecting patterns of seed dispersal to seedling establishment. Journal of Ecology, 88, 622 – 633. Roth I (1977) Fruits of angiosperms. In: Encyclopedia of Plant Anatomy (ed. Linsbauer K), pp. 374 – 381. Gebrüder Borntraeger, Berlin. Scheiner S, Roessli D, Excoffier L (2000) ARLEQUIN, version 2.0. A Software for Population Genetics Data Analysis. Genetics and Biometry Laboratory, University of Geneva, Geneva, Switzerland. Schnabel A, Beerli P, Estoup A, Hillis D (1998a) A guide to software packages for data analysis in molecular ecology. Advances in Molecular Ecology, 306, 291– 296. Schnabel A, Nason JD, Hamrick JL (1998b) Understanding the population genetic structure of Gleditsia triacanthos L. seed dispersal and variation in female reproductive success. Molecular Ecology, 7, 819 – 832. Slate J, Marshall T, Pemberton J (2000) A retrospective assessment of the accuracy of the paternity inference program CERVUS. Molecular Ecology, 9, 801– 808. Sork VL, Nason J, Campbell DR, Fernández JF (1999) Landscape approaches to historical and contemporary gene flow in plants. Trends in Ecology and Evolution, 14, 224. Sosinski B, Gannavarapu M, Hager LD et al. (2000) Characterization of microsatellite markers in peach [Prunus persica (L.) Batsch]. Theoretical and Applied Genetics, 101, 421– 428. Tautz D (1989) Hypervariability of simple sequences as a general source for polymorphic DNA markers. Nucleic Acid Research, 17, 6463 – 6471. Ueno S, Tomaru N, Yoshimaru H, Manabe T, Yamamoto S (2000) Genetic structure of Camellia japonica L. in an old-growth ever- green forest, Tsushima, Japan. Molecular Ecology, 9, 647– 656. Valle F, Gómez F, Mota F, Díaz C (1989) Parque Natural de Cazorla, Segura y Las Villas. Guía botánico-ecológica. Editorial Rueda, Madrid. Wheelwright NT, Orians GH (1982) Seed dispersal by animals: contrasts with pollen dispersal, problems of terminology, and constraints on coevolution. American Naturalist, 119, 402 – 413.

These results are part of a broader study on the effect of animal frugivores on the genetics and demography of fleshy-fruited plants. Ongoing research includes the use of microsatellite markers in parentage analysis to study seed dispersal in relation to frugivore foraging. P. Jordano is interested in the evolutionary ecology of plant-animal mutualisms and the demographic and genetic effects of the interaction. J. A. Godoy is interested in the development and application of molecular techniques to studies of ecology and conservation of endangered species (individual, parentage, and population genetic analyses). They coordinate the Molecular Ecology Laboratory at the Estación Biológica de